14:42 <nomis> robmur01: I have a login prompt. Turns out that the numbering of the mmc devices (emmc/sdio/sdmmc) did not agree with the specified root device...

14:46 <robmur01> nomis: hooray! Now thanks to GPT, "PARTUUID=" is your friend

14:48 <nomis> for now I took the cheap route and disabled sdio, that fixed the mmcblk-numbering :]

14:48 <robmur01> ...at least until you get into the habit of cloning eMMC images to SD cards and back again, sometimes booting with both installed, and wondering why changes to files seem to go missing...

14:49 <nomis> heh :)

15:16 <urjaman> robmur01: you can use PARTUUID with MBR partitioning too (well it's the MBR disk id and partition number then, but still)

15:17 <robmur01> urjaman: sure, I guess what I was trying to say is that GPT gives you no excuse for *not* using it ;)

15:18 <robmur01> (or at least UUIDs in general; I still don't really understand what the deal with both UUID and PARTUUID is...)

15:32 <JPEW> robmur01: IIRC one is a property of the filesystem (e.g. part of the superblock) and one is a "property" of the partition

15:34 <robmur01> JPEW: ah, that sounds reasonable, and like something I may have had a vague grasp on at one point, thanks!

15:38 <mps> can the arm-trusted-firmware (ATF) be built with arm64 compiler or it must be built with arm32?

15:39 <robmur01> heh, indeed /dev/disks/by-* clearly demonstrates that my machine has both partitions that don't contain filesystems, and filesystems that aren't on partitions

15:40 <robmur01> mps: ATF for 64-bit targets needs an aarch64 toolchain; for 32-bit targets it needs a 32-bit one

15:41 <robmur01> the fiddly part is that (64-bit) RK3399 ATF also contains a bit of 32-bit firmware for the Cortex-M0 cores, so wants both

15:42 <mps> robmur01: thanks. but is Cortex-M0 used on aarch64/rk3399

15:43 <mps> I'm trying to build ATF for aarch64 on alpine linux

15:43 <robmur01> right, for that you need a separate 32-bit toolchain (unless you can play tricks with Clang)

15:43 <robmur01> with GCC, aarch64 and arm are entirely separate backend targets

15:44 <robmur01> so there's nothing like "-m32" on x86

15:44 <mps> on alpine we have gcc-arm-none-eabi, and I can build ATF with it. but would be nice if it can be built without gcc-arm-none-eabi

15:45 <mps> and I'm building on aarch64 box or lxc container

15:45 <robmur01> does ATF allow to omit building the M0 firmware? IIRC it doesn't actually do much, so the machine may be able to survive without it

15:46 <mps> aha, ok. will look at this

16:11 <wens> dwc2 ok rk3328 is really weird in u-boot... if it probes before the ehci/ohci host pair, it can't find connected devices; otherwise it can :/

16:32 <robmur01> wens: I'd suspect something phy-related

16:36 <robmur01> similar to how it happens to sort-of-work in Linux until you unplug for the first time